geckordp

This is a client implementation of Firefox DevTools over remote debug protocol in python.

It essentially exposes the raw api to interact with the debug server and has some similarities with a common webdriver. See also Documentation.

What’s possible with geckordp?

Geckordp is meant to be used as a low level library to build tools on top. With a few helpers like the WebExtension-API and a proxy server, it can be feature rich enough for:

web ui-testing
extension testing
browser test tools
webdriver
data scraping
https recording
network traffic analysis
remote controller for browser
…and possibly more

Getting Started

To use Geckordp, install it with:

pip install geckordp
# python -m pip install geckordp
# python -m pip install geckordp[develop]

Alternative installation with venv:

python -m venv .venv
source .venv/bin/activate
python -m pip install geckordp

Documentation can be generated with:

sphinx-build -a -c docs/src -b html docs/build docs

Package signature can be checked with:

pip download --no-deps geckordp
wget https://github.com/jpramosi.gpg -O pub.gpg
wget https://raw.githubusercontent.com/jpramosi/geckordp/master/signatures/geckordp-latest.zip.asc -O latest.asc
gpg --no-default-keyring --output pub.sig --dearmor pub.gpg
gpg --no-default-keyring --keyring ./pub.sig --verify latest.asc geckordp-*.zip

# exemplary output:
gpg: Signature made So 23 Okt 2022 14:08:20 CEST
gpg:                using RSA key 21F942661941E642894267539B8551A5AEA1227A
gpg:                issuer "jimmy.pramosi@protonmail.com"
gpg: Good signature from "Jimmy Pramosi (git) <jimmy.pramosi@protonmail.com>" [ultimate]

Usage

import json

from geckordp.actors.root import RootActor
from geckordp.firefox import Firefox
from geckordp.profile import ProfileManager
from geckordp.rdp_client import RDPClient

""" Uncomment to enable debug output
"""
# from geckordp.settings import GECKORDP
# GECKORDP.DEBUG = 1
# GECKORDP.DEBUG_REQUEST = 1
# GECKORDP.DEBUG_RESPONSE = 1


def main():
    # clone default profile to 'geckordp'
    pm = ProfileManager()
    profile_name = "geckordp"
    port = 6000
    pm.clone("default-release", profile_name)
    profile = pm.get_profile_by_name(profile_name)
    profile.set_required_configs()

    # start firefox with specified profile
    Firefox.start("https://example.com/", port, profile_name, ["-headless"])

    # create client and connect to firefox
    client = RDPClient()
    client.connect("localhost", port)

    # initialize root
    root = RootActor(client)

    # get a list of tabs
    tabs = root.list_tabs()
    print(json.dumps(tabs, indent=2))

    input()


if __name__ == "__main__":
    main()

Tested Platforms

Tested Platform	Working	Firefox-Version	Geckordp-Version
Windows (x64)	?	129.0	1.0.1
Ubuntu 24.04	yes	129.0	1.0.1
macOS 12	?	129.0	1.0.1

Geckordp requires minimum Python 3.10 and the latest Firefox build. Older versions of Firefox may also work as long the API changes are not too drastically. In case of doubt, clone and run tests with:

cd <your-repositories-path>
git clone https://github.com/jpramosi/geckordp
cd geckordp
python -m venv .venv
source .venv/bin/activate
python -m pip install -e $PWD
pytest tests/ &> test.log

Versions

Geckordp is following the semantic versioning (Major.Minor.Patch).
Bug fixes not affecting the API increment the patch version, backward compatible API additions/changes increment the minor version, and backward incompatible API changes increment the major version. Breaking changes will be additionally labeled with [breaking] in the changelog.
Older versions of Geckordp with its corresponding Firefox version can be found here. But keep in mind it may have missing actors or bug-fixes.

Contribute

Every help in form of issues, questions or pull requests are very appreciated. If you would like to improve the project there are a few things to keep in mind:

For submitted code:

formatting
tests (optional)
should basically reflect the geckodriver api (if possible)

Even incomplete contributions are welcome and can be finished together.
For problems or improvements just open an issue.

Eventually you can also contribute to the project just by asking what do you need (examples, a specific task, features or whether something is feasible) on the issue tracker. Often it will also help other users too.

Develop

To get an idea what’s missing, here is a rough list of some notable objectives:

add remaining actors from geckodriver
add documentation for all actors its functions (even official repository got none)

If you are willing to get your hands dirty, please follow me here.

Technical Details

To be able to communicate with the server, a pre-configured profile is required.

Geckordp offers additional helper functions to resolve this problem with the ProfileManager.

The following flags are changed on profile configuration:

### disable crash-recover after 'ungraceful' process termination
("browser.sessionstore.resume_from_crash", False)

### disable safe-mode after 'ungraceful' process termination
("browser.sessionstore.max_resumed_crashes", 0)
("toolkit.startup.max_resumed_crashes", -1)
("browser.sessionstore.restore_on_demand", False)
("browser.sessionstore.restore_tabs_lazily", False)

### set download folder (not set by firefox)
("browser.download.dir", str(Path.home()))

### enable compatibility
("devtools.chrome.enabled", True)

### don't open dialog to accept connections from client
("devtools.debugger.prompt-connection", False)

### enable remote debugging
("devtools.debugger.remote-enabled", True)

### allow tab isolation (for e.g. separate cookie-jar)
("privacy.userContext.enabled", True)

### misc
("devtools.cache.disabled", True)
("browser.aboutConfig.showWarning", False)
("browser.tabs.warnOnClose", False)
("browser.tabs.warnOnCloseOtherTabs", False)
("browser.shell.skipDefaultBrowserCheckOnFirstRun", True)
("pdfjs.firstRun", True)
("doh-rollout.doneFirstRun", True)
("browser.startup.firstrunSkipsHomepage", True)
("browser.tabs.warnOnOpen", False)
("browser.warnOnQuit", False)
("toolkit.telemetry.reportingpolicy.firstRun", False)
("trailhead.firstrun.didSeeAboutWelcome", True)

Once the new profile was created, Firefox can be started with it. However, actors need to be initialized at first.

Some actors need to call additional functions to get initialized on server-side. But this is not always necessary and depends on what is actually needed. These required functions and its actors are initialized respectively used in this order according to the pcap-dumps.

| Browser initialization:

RDPClient()                 -> .connect()
    v
RootActor()                 -> .get_root()
    v
DeviceActor()               -> .get_description()
    v
ProcessActor()              -> .get_target()
    v
WebConsoleActor()           -> .start_listeners([])
    v
ContentProcessActor()       -> .list_workers()

| Tab initialization:

TabActor()                  -> .get_target()*
    v
WebConsoleActor()           -> .start_listeners([])*
    v
ThreadActor()               -> .attach()*
    v
WatcherActor()              -> .watch_resources(...)*
    v
TargetConfigurationActor()

*required if this actor will be used or events are wanted

The following hierarchy diagram shows dependencies between the actors and how to initialize individual actors: diagram

For debugging purposes, Geckordp can be configured to print out requests and responses to better understand the structure of the json packets. To enable it use:

from geckordp.settings import GECKORDP
GECKORDP.DEBUG = 1
GECKORDP.DEBUG_REQUEST = 1
GECKORDP.DEBUG_RESPONSE = 1
# environment variables can also be used for e.g.
# GECKORDP_DEBUG_RESPONSE=1

Other noteworthy general hints, issues or experiences:

actor initialization (plus the related functions like attach, watch or listening) on blank new tabs may get detached after visiting a new url and must be reinitiated (can be avoided if the page got a html header & body)
received messages are just plain python dictionaries and most of the time it has consistent fields which can be directly accessed
failed requests will return ‘None’
actors can have multiple contexts, that means different actor IDs can have the same actor model (for e.g. WebConsoleActor for process or tab)
called functions within manually registered async handlers on RDPClient can not call functions which emitting ‘RDPClient.send_receive()’ later in its execution path (instead use non-async handlers in this case)
on a new Firefox update it can happen that a few events doesn’t get caught by the RDPClient handler or requests getting a wrong response, unfortunately a few event/response packets doesn’t follow the same pattern and events must be manually specified in Geckordp which can have the implied side effects

License

MIT License

Copyright (c) 2024 jpramosi

Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
"Software"), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:

The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.